On the road to safe and sustainable mobility, we are facing a huge dilemma. On the one hand, the expansion of automotive semiconductors is increasing the significant demand for electricity; on the other hand, the world is struggling to achieve sustainable energy development. So how should the growth of automotive semiconductors be organically combined with sustainable development?
Microelectromechanical systems (MEMS) and sensors play an important role in ensuring the practicality and safety of autonomous vehicles, and their latest developments are critical to achieving sustainability.
How do current trends in the automotive world relate to the era?
In the past 10 years, MEMS devices have moved from that offline era (just implementing certain product features or functions) into a more powerful “online” era. Connecting sensors to the cloud in the online age enables performance improvements and technology convergence, which are key to making sensor information available to any ecosystem.
Take the car as an example, from direct airbag deployment to simultaneously calling emergency services and uploading relevant information about the impact and the location of the accident vehicle to assist rescue operations.
In terms of mobile technology, people now need a sustainable “online life” era, which has two characteristics. The first is people-centric, improving the way people interact with the world around them. Such technologies are safe and non-intrusive while enabling the operation of a driver’s assistant. The second is sustainable development, protecting the planet on which people live.
Passenger vehicles are responsible for approximately 3 billion tons of CO2 emissions globally. Electrification has a central role to play in dramatically reducing these emissions. At the same time, we will also see a significant increase in vehicles with increasing levels of automation and, eventually, autonomous driving.
Inertial sensors for autonomous vehicles: Smart, safe and precise
Smart inertial sensors are critical to higher levels of autonomous driving. According to the definition of SAE International, these higher levels of autonomous driving are L3, L4 and L5 respectively. Compliance with strict safety standards is critical to protect vehicle occupants, pedestrians and other traffic participants.
The attribute of intelligence is very important: a self-driving car needs to be able to react to every possible situation. They must be powered by AI algorithms that mimic (or even improve) human behavior and reaction times.
By integrating processing power directly within the sensor, these algorithms can directly analyze sensor data and perform operations without the need to transmit large amounts of sensor data to a host computer or cloud for processing. This not only speeds up the reaction time, but also greatly reduces the system power consumption.
Proper programming of elements in the controller on the sensor then ensures that the sensor is human-centric and allows the vehicle to implement self-configurable solutions through hard artificial intelligence while optimizing the power budget.
Smart inertial sensors must also be safe. Self-driving cars must meet the strictest safety standards. Smart inertial sensors should help self-driving cars accurately read their surroundings, since the car must know where it is, where it is going, and the orientation of all other vehicles around it. It’s not just other vehicles to keep an eye on, a self-driving car must also know where each obstacle is, because that’s just as important to driving safely.
Nowadays, cars have integrated more and more embedded circuits, which are functional safety systems implemented in hardware to improve power efficiency. Inertial sensors serve many purposes, such as compensating for camera images affected by tilt and vibrations caused by steering maneuvers and road noise. These types of systems typically require Automotive Safety Integrity Level B (ASIL-B) certification.
On the other hand, systems used for autonomous driving face stricter requirements, requiring e.g. ASIL-D certification. The next generation of inertial sensors will be designed with this in mind, possibly with an independently tested software library onboard the vehicle to facilitate safety certification to ASIL-B and above.
In addition, the sensors must be stable, reliable and able to withstand harsh environmental conditions. They must also be secure to prevent unauthorized access and data leakage.
Finally, smart inertial sensors must be precise, as self-driving cars require precise and accurate data to operate safely. It is required that these vehicles have a driving deviation of less than 0.1 degrees during one hour of continuous driving, and complete safe parking and flameout with an absolute accuracy of 20cm.
This accuracy is already comparable to that of lunar navigation systems, yet it is now being achieved with ordinary components in standard inertial sensing packages. It is indeed unbelievable!
Ensuring the accuracy of the data helps reduce the workload of the application processing, thereby reducing power consumption by minimizing the training demand on the data.
Note also that time delay affects accuracy: no matter how high the sensitivity, or how deep the resolution, the environment is constantly changing, so the information is not accurate in the first place. Therefore, low latency is an important attribute of smart sensors.
The path to fully autonomous vehicles is on the way, though the destination is not yet obvious. But there is no doubt that sustainable autonomous vehicles are on the highway to the future of mankind, and the main challenge will still be sustainable autonomous driving. It is also clear that smart, safe, and accurate inertial sensors are critical, and at least from a sensor development perspective, there is a clearer picture.
In recent years, state-of-the-art MEMS have enabled ADAS applications that require high-precision sensing capabilities, although this has little relevance to safety application development. On the other hand, the push by OEMs for advanced safety systems has spurred rapid development of safety applications with limited precision requirements for sensing capabilities.
Now, if we really want to pave the way for the sensorization of future sustainable vehicles, two needs need to be integrated: safety on the one hand and precision on the other. One must not forget that everything requires absolute intelligence, with self-configurable sensors, and low-power optimized systems for data processing. And these will ultimately drive automakers to provide global solutions.